How to check if initial conditions are unbalanced or not?

Report or discuss software problems and other woes

Moderators: arango, robertson

Post Reply
Message
Author
jayasankar
Posts: 15
Joined: Fri Sep 20, 2013 1:09 pm
Location: India

How to check if initial conditions are unbalanced or not?

#1 Post by jayasankar » Tue Nov 22, 2016 9:43 am

Hi all,
I have been trying to setup a model and it is resulting in BLOW UP in like 10-30 steps depending up on grid configuration and time step. I guess it is because of the initial conditions, temperature and salinity..

How to see if they are unbalanced or not, and how to correct the same?
Could it be solved by smoothing the initial data?

Thanks in advance.

jzavala
Posts: 7
Joined: Sat Jan 28, 2006 1:58 am
Location: IMCS

Re: How to check if initial conditions are unbalanced or not

#2 Post by jzavala » Wed Nov 23, 2016 4:07 pm

Jayasankar,
I assume that you are also providing the other variables in the IC file (velocities and sea surface elevation).
If you are taking your IC from another model they should be mostly balanced unless the geometry of your ROMS application is very different to that of the other model.
You should also check that you don't have any NaNs in your IC and forcing files.
The best practice will be to output your solution every time step and see where it blows up and try to find why. You might find for instance that your land-se mask is bogus there, or that your IC interpolated to the ROMS grid is really unrealistic there.
Also, if you post your log file it might be possible to spot other problems such as the use of a too large DT, etc.

Jvr

jayasankar
Posts: 15
Joined: Fri Sep 20, 2013 1:09 pm
Location: India

Re: How to check if initial conditions are unbalanced or not

#3 Post by jayasankar » Wed Nov 23, 2016 10:27 pm

Hi Jvr,

Thanks a lot for the prompt reply.
yes the IC file contains all needed data. The IC file is made by interpolating SODA data.
So as you said it could probably be the geometry of my model.

The model was instable and I found that the density was rising through out the domain.
So I tried to smoothen the temperature and salinity values in the IC file using matlab function 'smooth3'.

I have noticed that the more I smoothened (by changing the convolution kernel of smooth3 function) the lower was the density rise rate and model ran longer. The BLOW UP location also changed.

I will check for NaN values, meanwhile I would be grateful if you could give more insights.
I have tried to upload the .out file but somehow the upload is failing.

Jayasankar

Post Reply